MortarBench - Transaction IDs: leaderboard
The transaction-list questions of MortarBench (Toles et al., 2026), such as which deposits count as large deposits, on synthetic mortgage files, each a ULAD loan application and a bank statement, with questions drawn from a production loan-officer assistant. Set-overlap F1 in percent between the returned and gold transaction IDs, mean of three trials; higher is better.
Metric: F1 (%; the 108 transaction-list cases; mean of 3 trials over MortarBench's 188 synthetic mortgage files (47 real loan-officer questions x 4 mutated profiles, each a ULAD application plus a bank statement); Boolean answers score 1 or 0 and transaction or account lists get set-overlap F1; baseline agent without the authors' CRIT confidence filter, temperature 1 (GPT-5.5 at its fixed default)). Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 79.7 | #181 |
| 2 | GPT-5.5 | 72.3 | #48 |
| 3 | Gemini 3.1 Pro (Preview) | 72.2 | #91 |
| 4 | Kimi K2.6 | 63.4 | #173 |
| 5 | Claude Sonnet 4.6 | 63 | #158 |
| 6 | Qwen 3 32B | 54.8 | #960 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mortarbench-transaction-ids · How It Works · Data refreshed daily, snapshot 2026-11-12.